A Fast Fuzzy DRL-Based Joint Beam Design and Power Allocation for Multi-Beam GEO-LEO Coexisting Satellite Networks

As demand for ubiquitous connectivity grows, integrating satellite communications into sixth-generation (6G) networks has emerged as a crucial strategy to enhance global coverage, especially in remote and underserved regions. However, achieving the stringent performance, reliability, and spectral efficiency required for 6G presents significant challenges. Coexisting geostationary (GEO) and low Earth orbit (LEO) satellite networks offer a promising solution by enabling complementary coverage and enhanced service capabilities. Nonetheless, a critical challenge is managing intersystem interference from the LEO satellite system on the GEO system when sharing spectral resources, all while maintaining the performance of both systems. To address this, this paper introduces a fast fuzzy deep reinforcement learning (DRL)-based approach for joint beam design and power allocation in multi-beam GEO-LEO coexisting satellite networks. A robust design problem of LEO beam size and power allocation is formulated to maximize the spectral efficiency of the LEO system, considering tolerable interference on the GEO system, frequency reuse schemes employed by both GEO and LEO systems, and Doppler frequency offset induced by LEO satellite movement. A fast DRL algorithm, integrating fuzzy logic, post-decision state, and deep deterministic policy gradient, is proposed to solve this problem. Numerical results demonstrate a faster learning convergence rate for the proposed DRL algorithm compared to benchmark algorithms and confirm that the proposed method enhances LEO spectral efficiency while maintaining tolerable intersystem interference on the GEO system.